MyMEPS practice data

This time we focus on Total Expenditures in 2014 along with a persons insurance coverage. We use INSCOV14 for the insurance variable

a) Using the H171 data, extract these two variables into a new data set, take care of coding for missing data, and labeling the categorical data. Create the variable for expenditures to indicate whether a person has exactly zero expenditures or positive expenditures.

b) For the comparison between the indicator expenditure variable and insurance, create 3 tables: one that calculates the cell percentages (so that all entries add to 1), one that calculates row percentages (so that entries by row sum to 1), and one that calculate column percentages (so that entries by column sum to 1).

Interpret the summaries in each table.

c) Provide a numerical descriptive summary of expenditures in total by insurance category using the psych package and summarize the results verbally.

d)Discuss why a t-test using INSCOV14 is not appropriate in question (3) to assess the statistical significance of the difference of total expenditures by insurance category.

e)Conduct a 22-test between the indicator variable for total expenditures and insurance. Be sure to state clearly your null and alternative hypotheses, along with the conclusion.

f)Provide a numerical descriptive summary of expenditures of those with positive expenditures by insurance category using the psych package and summarize the results verbally. (Hint: you can subset those with positive expenditures to a new data frame.)

g)Create a boxplot of positive expenditures by insurance category, not including the outliers so that you can better see the data. Verbally interpret your results (and the usefulness of the plot) and compare the plots versus your summary statistics in question (3).

h)One common way of dealing with highly skewed data (like Expenditures) is by taking a natural log transformation. You may recall that the log of a lognormally distributed random variable is normally distributed. Thus researchers take a log of a variable to see if it helps normalize it.

i)Create a new variable that is the log of positive Expenditures. (Things get messed up if you keep the zero expenditures in the computation.) Compare histograms of the original distribution of Expenditures versus the log of Expenditures. Discuss whether the new distribution is more normally distributed.